ScrambleToolBench - Base - Tasks Solved: leaderboard
Metric: Average tasks solved per five-task episode (0-5). Source: arxiv.org. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 5 |
| 2 | Gemini 3.5 Flash | 5 |
| 3 | Claude Sonnet 5 | 5 |
| 4 | Claude Sonnet 4.5 | 4.5 |
| 5 | Qwen 3.6 27B | 4 |
| 6 | Gemini 2.5 Pro | 3.8 |
| 7 | Gemini 2.5 Flash | 2.05 |
| 8 | GPT-5.4 | 2.05 |
| 9 | Gemma 4 31B | 1.5 |
| 10 | Qwen 3.6 35B A3B | 1.2 |
| 11 | Gemma 4 26B A4B | 0.4 |
| 12 | GPT-OSS-120B | 0.05 |
| 13 | Gemini 3.1 Flash Lite | 0.05 |
| 14 | GPT-5.4 Mini | 0.05 |
| 15 | GPT-OSS-20B | 0 |
Interactive version: theaggregate.ai/benchmark?slug=scrambletoolbench-base-tasks-solved · How It Works · Data refreshed daily, snapshot 2026-09-19.